Introduction: Why Code for Data Science?
Phil Chodrow
Tuesday, August 27th, 2019
Some Things That Aren’t Data Science
The Cloud\(^{\mathrm{TM}}\)
BIG DATA!!1!!
Data Science Is:
- Gathering data that matters.
- Asking questions that matter about your data.
- Choosing appropriate methods to answer those questions.
- Implementing solutions that meet stakeholder needs.
You Can Do Data Science With:
- A pencil and paper.
- A calculator.
- Excel.
- Coding:
R, Julia, Python….
Why Not Excel?
- Flexibility:
- Limited statistics/ML
- Poor visualizations
- Reproducibility:
- Platform limited
- Can’t version control
- Difficult to inspect code/troubleshoot
- Scalability:
- Excel can handle ~1M rows at best
Why Code?
- Flexibility:
- “There’s a package for that”
- Custom analysis and visualization
- Reproducibility:
- Cross-platform, often FOSS
- Version-control with
git
- Easy to inspect code
- Scalability:
R and python: ~10M rows easily on a laptop.
Version Control with git
- Break your workflow into manageable stages; easily collaborate; access cool code.
- Promote your brand: share your work, build a portfolio, host your website.
- Used at: Google, Facebook, Netflix, Amazon, Apple, Twitter, Microsoft… (source)
Data Analysis with R
R is the best language in the world for learning data science.R is one of the best languages in the world for doing data science.R tends to be preferred in academia and among “statisticians,” while python is more popular among “computer scientists” and “data scientists”- Most practicing data scientists know and use both.
Optimization with Julia and JuMP
Julia is high-performance, open-source dynamic language for technical computing – easy writing, fast compute times.- Developed at MIT.
JuMP is a package for optimization in Julia – developed by ORC students!- Not everyone uses
Julia…yet.
…yes, there will be an opportunity to learn Python later in the semester.
What can you pick up in two days?
- You are not going to become an expert in two days.
- But…
- You will know the basic concepts and vocabulary of data science – enough to employ the most important skill of all.
The most important skill of all…
The most important skill of all…
Gameplan
- Today: Version Control, Basic Data Analysis and Visualization in
R, RMarkdown.
- Tomorrow: Optimization in
Julia and JuMP, selected presentations.
- Both days: mini-project, partner work, lots of exercises.
Exercise 0
- Look left.
- Look right.
- Pick a partner (groups of 3 are fine).
- Give them a professional, yet friendly smile.
- You are going to need them soon.
